SVD-VMD algorithm and its application in leak detection of natural gas pipeline
Hong Li, L. Chu, Jingyi Lu, Qingqiang Liu, Fu Li, Kun Zhang
Abstract
In the study of pipeline leak detection, it is arduous to identify the expected target accurately because of the influence of various noises in practicable application. In order to realize the effective identification of pipeline signals under different working conditions. An integrated leak detection method based on singular value decomposition (SVD) and variational mode decomposition (VMD), as well as probabilistic neural network (PNN), that is SVD-VMD-PNN, is proposed. First, a novel index based on the relative change rate of singular value energy (SVE) is presented to determine the reconstructed order of singular values effectively. Second, the optimal K and α can be selected adaptively by using the arithmetic optimization algorithm (AOA). The ratio of the mean and variance of the dispersion entropy (DE) is used as the fitness function of AOA. Subsequently, the feature vectors based on the energy entropy of each sub-mode for leak detection are extracted. Finally, PNN is employed for leak detection. The laboratory experiments classification results indicate that the proposed SVD-VMD-PNN method can achieve high accuracy of 100% to identify the leak signal and non-leak signal, it is further verified that the proposed method is feasible and superior to other methods.